service gurad

[Shaik Hyder Shakeel]

what is steps by details need at SG before upgrade oracle 9i as halt package is recomended or no

[Duncan Edmonstone]

This is so dependent on your own configuration its almost impossible to say...

do you share executables between cluster nodes, or do you have local oracle binaries on each system?

do you monitor oracle processes?

Also when you say upgrade do you mean from or to 9i, and what are you upgrading from/to?

As usual in these forums, the metaphor here is you've gone to a mechanic and said - "I need a new engine for my vehicle" - but who knows what you're driving??

HTH

Duncan

[Steven E. Protter]

Shalom,

For 9i RAC
http://www.docs.hp.com/en/T1859-90038/ch03s14.html

In general, there is a lot of commonality to doing ServiceGuard and RAC.

SEP

[Shaik Hyder Shakeel]

it is not on every local node but in share , and i haven't monitor tools. ined to upgrade from 9i to 10g so that ineed to whay i can do regarding MCSG before start upgrade oracle by more details

[Sajjad Sahir]

dear Shaik

please give me cmviewcl -v output
i think u have two node cluster with slvm

please send more details


Sajjad Sahir

[Duncan Edmonstone]

so when you stop oracle manually (i.e. via a shutdown immediate), does it cause the package to failover?

if not, then you can probably just go ahead and upgrade to 10g without doing anything to Serviceguard.

The only thing I'd watch for is where the Oracle inventory pointer (OraInst.loc) for the OUI is... if Oracle wasn't installed by an experienced DBA, then chances are it is just in /var/opt/oracle on just one of the nodes. If it is then you'll need to do the upgrade on that node... if its somewhere else on one of your clustered filesystems you'll need to use the -invPtrLoc option on the runInstaller command line during the upgrade.

HTH

Duncan

[Shaik Hyder Shakeel]

my question i need halt package on sg bfeore start upgrade oracle 9i to 10g . if yes tell me steps to do that

[Duncan Edmonstone]

and I answered that - if you can stop Oracle manually without causing a cluster failover then NO you don't need to halt the package.

And if you don't need to halt the package, you don't need any instructions do you?

Duncan

[Shaik Hyder Shakeel]

thanks for your fast replay but after upgrade the home file will be change so i need to reconfigure package. what i can do?

[Duncan Edmonstone]

well now we're taling specifics of your environment... aboou which you bhave told us nothing...

so I'mn assuming that your last comment meant that after the upgrade your ORACLE_HOME will have changed to a new location yes?

assuming that is what you meant my next questions are:

- is that ORACLE_HOME on a new filesystem, or one of the existing filesystems in the cluster? If on a new filesystem, is that

- are you sing the Enterprise Cluster Master Toolkit Oracle integration and if so waht version. Or has someone "rolled their own" cluster integration for Oracle?


- can you post the output of:

cmgetconf -p

replacing with the serviceguard package name of your oracle instance

HTH

Duncan

[Duncan Edmonstone]

gaaah real finger trouble this morning, let me post that again:

well now we're talking specifics of your environment... about which you have told us nothing...

so I'm assuming that your last comment meant that after the upgrade your ORACLE_HOME will have changed to a new location yes?

assuming that is what you meant my next questions are:

- is that ORACLE_HOME on a new filesystem, or one of the existing filesystems in the cluster? If on a new filesystem, is that in the same VG as existing cluster filesystems or in an entirely new VG?

- are you sing the Enterprise Cluster Master Toolkit Oracle integration and if so waht version. Or has someone "rolled their own" cluster integration for Oracle?


- can you post the output of:

cmgetconf -p

replacing with the serviceguard package name of your oracle instance

HTH

Duncan

[Shaik Hyder Shakeel]

the same file system but new path for home file after upgrade . i know need to change some paramter at control file of packge as custom define run and halting the steps are:
1- stop package manualy not halting package
2- make upgrade oracle so that new path for home file at the same file system
3- reconfigure packge (control file)
4- ckeck package configuration
5-apply configuration of cluster package
please ckeck this steps and tell me the correct configuration if false

[Viktor Balogh]

if you only modify the cntl file, (and not the conf file)
you shouldn't apply the package configuration again.

[Shaik Hyder Shakeel]

i need halting package when reconfigure control file of package

[Shaik Hyder Shakeel]

output of cmviewcl -v
CLUSTER STATUS
sjodbcl up

NODE STATUS STATE
sjodbn1 up running

Cluster_Lock_LVM:
VOLUME_GROUP PHYSICAL_VOLUME STATUS
/dev/vgora /dev/dsk/c11t0d1 up

Network_Parameters:
INTERFACE STATUS PATH NAME
PRIMARY up 0/6/1/0 lan3
PRIMARY up 0/4/2/0 lan1
STANDBY up 0/2/1/0 lan4

PACKAGE STATUS STATE AUTO_RUN NODE
oradbpkg up running enabled sjodbn1

Policy_Parameters:
POLICY_NAME CONFIGURED_VALUE
Failover configured_node
Failback manual

Script_Parameters:
ITEM STATUS MAX_RESTARTS RESTARTS NAME
Service unknown 0 0 oracle_db
Subnet up 10.196.7.0

Node_Switching_Parameters:
NODE_TYPE STATUS SWITCHING NAME
Primary up enabled sjodbn1 (current)
Alternate up enabled sjodbn2

NODE STATUS STATE
sjodbn2 up running

Cluster_Lock_LVM:
VOLUME_GROUP PHYSICAL_VOLUME STATUS
/dev/vgora /dev/dsk/c11t0d1 up

Network_Parameters:
INTERFACE STATUS PATH NAME
PRIMARY up 0/6/1/0 lan3
PRIMARY up 0/4/2/0 lan1
STANDBY up 0/2/1/0 lan4

[Duncan Edmonstone]

if you keep asking the same question in multiple threads without posting the requested material, you are really never going to get a consistent answer. So stop opening new threads in which you repeat the same question and answer this which I asked before:

- are you sing the Enterprise Cluster Master Toolkit Oracle integration and if so what version. Or has someone "rolled their own" cluster integration for Oracle?

- if you're not sure we might be able to figure this out... the first stage of figuring this out is to post:

- the output of "cmgetconf -p oradbpkg"

- an "ls" of the packages directory (which presumably is /etc/cmcluster/oradbpkg)

The point is - this isn't Windows where the cluster configuration for an application is ususally restrictively consistent between different implementations - there are a dozen different ways to configure Oracle into a Serviceguard package (which is _good_ because it makes UNIX more flexible) and we don't know which way yours is configured so its very difficult to give advice...

HTH

Duncan

[Shaik Hyder Shakeel]

this output of ls my directory package
DBN1)/etc/cmcluster/oradbpkg $ll
total 544
-rwx------ 1 root sys 56842 Feb 14 2008 oradbpkg.cntl
-rwx------ 1 root sys 56842 Feb 14 2008 oradbpkg.cntl.bkg
-rw-r--r-- 1 root root 145889 Mar 26 19:24 oradbpkg.cntl.log
-rwx------ 1 root sys 16329 Aug 25 2007 oradbpkg.conf
attached is output of cmgetconf -p package
please i need steps to can make safe upgrade for oracle from 9i to 10g regarding to sg

[Duncan Edmonstone]

I know what you need but we need to know more about your environment...

So now look in the file oradbpkg.cntl

and find the section that looks like this:

# START OF CUSTOMER DEFINED FUNCTIONS

# This function is a place holder for customer define functions.
# You should define all actions you want to happen here, before the service is
# started. You can create as many functions as you need.

function customer_defined_run_cmds
{

...

}


I'm interested in what is in between the curly braces after the line:

function customer_defined_run_cmds

post that here - if it references another shell script, post the shell script here as well.

HTH

Duncan

[Shaik Hyder Shakeel]

in this run custome define path to start service of oracle and located at home file , this path will be change after upgrade
so i need sequense of upgrade oracle by comand line

[Duncan Edmonstone]

so you're saying that the actual code to start Oracle is in the customer_defined_run_cmds function rather than

sourced from another script? It would be easier for me to follow if you just posted it...


we also need to see the lines taht read something like:

SERVICE_NAME[0]="oracle_db"
SERVICE_CMD[0]="..."
SERVICE_RESTART[0]="..."

again it would be easier if you just posted the whole file...

HTH

Duncan

[Shaik Hyder Shakeel]

please tell me steps to can start upgrade . i think now you have all configuration needed at sg

[Duncan Edmonstone · Accepted solution]

So the code which starts/stops your database is then I'm assuming all in the script /u01/oracle/proddb/9.2.0/appsutil/scripts/PROD_sjodb/addbctl.sh

That's certainly not a HP or Oracle provided script, so must be intrrnslly developed... however in all likelihood it will just contain a "ORACLE_HOME=/u01/oracle/proddb/9.2.0" statement that you can update (you might want to relocate the script somehwere under your new ORACLE_HOME as well)

Assuming that is the case, and assuming as you said that 10g is just going to be installed on a directory in an existing filesystem in the cluster then as I originally said there is no need to halt the cluster package. You might want to after the upgrade to test out failover, but you won't need to for the upgrade...

The tasks for upgrade will basically be:

1) With cluster package UP Follow standard Oracle upgrade process from 9i to 10g as described in the excellent metalink article 316889.1

2) When you are happy that the database has been fully upgraded, edit the package control file and update both the customer_defined_run_cmds and customer_defined_halt_cmds section to reflect the new path if you did move the file /u01/oracle/proddb/9.2.0/appsutil/scripts/PROD_sjodb/addbctl.sh

3) distribute the updated package control file to all cluster nodes

4) Now if you want to check everything is working correctly, just halt the package and restart (on both nodes if you want) to prove your changes worked... you seem to have indicated you know the process for this, but just in case...

cmhaltpkg oradbpkg

cmrunpkg -n oradbpkg


----------------------------


of course if you don't want to move the file /u01/oracle/proddb/9.2.0/appsutil/scripts/PROD_sjodb/addbctl.sh and you've no intention of removing the old Oracle 9i home under /u01/oracle/proddb/9.2.0, the you really don't need to change anything at all in Serviceguard - you just need to update the addbctl.sh script with the new paths

HTH

Duncan

[Shaik Hyder Shakeel]

thanks to fast reply to me , my question i need to halt package before change control file or can change when package is running

[Duncan Edmonstone]

Yes, the control file can be changed when the package is running (well actually the part you are changing can be changed while the package is running - you can't add additional IP addresses or easily change filesystem mounts, but you're not going to do that)

Juts remember to distribute the updated file to all cluster nodes

HTH

Duncan

[Shaik Hyder Shakeel]

many thankd for you , you see all configuration at my enviroment so that you decise make upgrade without halting package